CIND 820 :- Stroke Prediction project¶

Table of Contents¶

  • Introduction
  • Data Exploration
  • Exploratory Data Analysis
  • Data Processing & Modeling
  • Conclusions

Introduction¶

  • The dataset to be used for this project are Stroke Prediction Dataset.
  • https://www.kaggle.com/datasets/fedesoriano/stroke-prediction-dataset
  • The dataset has about 5k records.
  • There are 12 columns: -id - gender - age - hypertension -heart_disease - ever married - work_type - Residence_type - avg_glucose_level - bmi -smoking_status – stroke.

Importing Libaries¶

In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

Data Exploration¶

In [2]:
df = pd.read_csv('healthcare-dataset-stroke-data.csv')
df.head()
Out[2]:
id gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status stroke
0 9046 Male 67.0 0 1 Yes Private Urban 228.69 36.6 formerly smoked 1
1 51676 Female 61.0 0 0 Yes Self-employed Rural 202.21 NaN never smoked 1
2 31112 Male 80.0 0 1 Yes Private Rural 105.92 32.5 never smoked 1
3 60182 Female 49.0 0 0 Yes Private Urban 171.23 34.4 smokes 1
4 1665 Female 79.0 1 0 Yes Self-employed Rural 174.12 24.0 never smoked 1
In [3]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5110 entries, 0 to 5109
Data columns (total 12 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   id                 5110 non-null   int64  
 1   gender             5110 non-null   object 
 2   age                5110 non-null   float64
 3   hypertension       5110 non-null   int64  
 4   heart_disease      5110 non-null   int64  
 5   ever_married       5110 non-null   object 
 6   work_type          5110 non-null   object 
 7   Residence_type     5110 non-null   object 
 8   avg_glucose_level  5110 non-null   float64
 9   bmi                4909 non-null   float64
 10  smoking_status     5110 non-null   object 
 11  stroke             5110 non-null   int64  
dtypes: float64(3), int64(4), object(5)
memory usage: 479.2+ KB
In [4]:
df.describe()
Out[4]:
id age hypertension heart_disease avg_glucose_level bmi stroke
count 5110.000000 5110.000000 5110.000000 5110.000000 5110.000000 4909.000000 5110.000000
mean 36517.829354 43.226614 0.097456 0.054012 106.147677 28.893237 0.048728
std 21161.721625 22.612647 0.296607 0.226063 45.283560 7.854067 0.215320
min 67.000000 0.080000 0.000000 0.000000 55.120000 10.300000 0.000000
25% 17741.250000 25.000000 0.000000 0.000000 77.245000 23.500000 0.000000
50% 36932.000000 45.000000 0.000000 0.000000 91.885000 28.100000 0.000000
75% 54682.000000 61.000000 0.000000 0.000000 114.090000 33.100000 0.000000
max 72940.000000 82.000000 1.000000 1.000000 271.740000 97.600000 1.000000
In [5]:
df.isnull().sum()
Out[5]:
id                     0
gender                 0
age                    0
hypertension           0
heart_disease          0
ever_married           0
work_type              0
Residence_type         0
avg_glucose_level      0
bmi                  201
smoking_status         0
stroke                 0
dtype: int64
In [6]:
df['stroke'].value_counts()
Out[6]:
0    4861
1     249
Name: stroke, dtype: int64

this dataset was highly unbalanced. Only 249 patients suffered a stroke while the remaining 4861 patients did not have the experience.

In [7]:
# Drop the id column
df.drop(columns=['id'], inplace=True)

This attribute has no other useful information and was only used to identify patients.

In [8]:
# Showing records where patient suffered from stroke but had missing value in bmi attribute.
df[df['bmi'].isna() & df['stroke'] == 1]
Out[8]:
gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status stroke
1 Female 61.00 0 0 Yes Self-employed Rural 202.21 NaN never smoked 1
8 Female 59.00 0 0 Yes Private Rural 76.15 NaN Unknown 1
13 Male 78.00 0 1 Yes Private Urban 219.84 NaN Unknown 1
19 Male 57.00 0 1 No Govt_job Urban 217.08 NaN Unknown 1
27 Male 58.00 0 0 Yes Private Rural 189.84 NaN Unknown 1
29 Male 59.00 0 0 Yes Private Rural 211.78 NaN formerly smoked 1
43 Female 63.00 0 0 Yes Private Urban 90.90 NaN formerly smoked 1
46 Female 75.00 0 1 No Self-employed Urban 109.78 NaN Unknown 1
50 Female 76.00 0 0 No Private Urban 89.96 NaN Unknown 1
51 Male 78.00 1 0 Yes Private Urban 75.32 NaN formerly smoked 1
54 Female 63.00 0 0 Yes Govt_job Urban 197.54 NaN never smoked 1
57 Male 78.00 0 0 Yes Private Urban 237.75 NaN formerly smoked 1
64 Male 75.00 0 0 Yes Private Urban 104.72 NaN Unknown 1
70 Female 76.00 0 0 Yes Govt_job Rural 62.57 NaN formerly smoked 1
78 Female 51.00 0 0 Yes Private Urban 165.31 NaN never smoked 1
81 Female 66.00 0 0 Yes Self-employed Urban 101.45 NaN Unknown 1
84 Male 58.00 0 0 Yes Private Urban 71.20 NaN Unknown 1
105 Male 58.00 0 0 Yes Private Urban 82.30 NaN smokes 1
112 Female 76.00 0 0 Yes Self-employed Urban 106.41 NaN formerly smoked 1
124 Female 72.00 0 0 Yes Private Urban 219.91 NaN Unknown 1
126 Male 78.00 1 0 Yes Self-employed Urban 93.13 NaN formerly smoked 1
129 Female 75.00 0 0 Yes Govt_job Urban 62.48 NaN Unknown 1
133 Female 38.00 0 0 Yes Private Rural 101.45 NaN formerly smoked 1
146 Male 65.00 0 0 Yes Self-employed Urban 68.43 NaN formerly smoked 1
150 Female 79.00 0 0 Yes Private Rural 169.67 NaN Unknown 1
160 Female 76.00 0 0 Yes Private Urban 57.92 NaN formerly smoked 1
161 Male 71.00 0 1 Yes Private Urban 81.76 NaN smokes 1
162 Female 1.32 0 0 No children Urban 70.37 NaN Unknown 1
167 Male 79.00 1 0 Yes Private Rural 75.02 NaN never smoked 1
170 Male 64.00 0 0 Yes Self-employed Rural 111.98 NaN formerly smoked 1
171 Female 79.00 1 1 No Self-employed Rural 60.94 NaN never smoked 1
174 Female 78.00 0 0 Yes Self-employed Rural 60.67 NaN formerly smoked 1
178 Female 80.00 0 0 Yes Govt_job Urban 110.66 NaN Unknown 1
183 Female 77.00 0 0 No Private Urban 81.32 NaN Unknown 1
189 Male 61.00 0 1 Yes Private Urban 209.86 NaN Unknown 1
198 Male 79.00 0 0 Yes Private Rural 114.77 NaN formerly smoked 1
200 Male 74.00 0 0 Yes Private Urban 167.13 NaN Unknown 1
218 Female 76.00 1 1 Yes Self-employed Urban 199.86 NaN smokes 1
227 Male 74.00 0 0 Yes Self-employed Rural 60.98 NaN never smoked 1
247 Male 71.00 1 0 Yes Self-employed Rural 87.80 NaN Unknown 1
In [9]:
# Replace the missing values with mean of bmi attribute
df['bmi'].fillna(np.round(df['bmi'].mean(), 1), inplace = True)

records with empty value in BMI was replaced with mean of BMI.

In [10]:
df = df[df['gender'] != 'Other']

only one record were categorized as ‘Other’, so it was deleted

Gender needs to be categorized as binary variable.

Exploratory Data Analysis¶

Univariate Analysis¶

In [11]:
numerical = ['age','avg_glucose_level', 'bmi']
df[numerical].hist(figsize=(8,6));
  • As seen in both skewness result and histograms, numerical features have skewness in different degrees.

Categorical Features¶

In [12]:
sns.countplot(data=df,x='gender')
Out[12]:
<AxesSubplot:xlabel='gender', ylabel='count'>

We have 2994 female and 2115 male

In [13]:
sns.countplot(data=df,x='hypertension')
Out[13]:
<AxesSubplot:xlabel='hypertension', ylabel='count'>
  • We have 498 patient with hypertension which represents at round 10 % of the sample.
In [14]:
sns.countplot(data=df,x='heart_disease')
Out[14]:
<AxesSubplot:xlabel='heart_disease', ylabel='count'>
  • We have 276 patient with heart disease which is 5.4 % of the sample.
In [15]:
sns.countplot(data=df,x='ever_married')
Out[15]:
<AxesSubplot:xlabel='ever_married', ylabel='count'>
  • 3353 people have been married and 1757 people are not married before.
In [16]:
sns.countplot(data=df,x='work_type')
Out[16]:
<AxesSubplot:xlabel='work_type', ylabel='count'>
  • 2925 people work in the private sector.
  • 819 people are self-employed
  • 657 people work at the government job.
In [17]:
sns.countplot(data=df,x='smoking_status')
Out[17]:
<AxesSubplot:xlabel='smoking_status', ylabel='count'>
  • 1892 people are never smoked , 789 people smoke

Bivariate Analysis¶

In [18]:
# Gender & Stroke
sns.countplot(data=df,x='gender',hue='stroke')
Out[18]:
<AxesSubplot:xlabel='gender', ylabel='count'>

there is not much difference between stroke rate concerning gender

In [19]:
# Hypertension & Stroke
sns.countplot(data=df,x='hypertension',hue='stroke')
Out[19]:
<AxesSubplot:xlabel='hypertension', ylabel='count'>

Well, young people rarely have hypertension, while the elderly frequently do. A stroke can be brought on by hypertension. Our data do not paint a very clear picture of hypertension. On patients with hypertension, there is not a lot of information.

In [20]:
# Married & Stroke
sns.countplot(data=df,x='ever_married',hue='stroke')
Out[20]:
<AxesSubplot:xlabel='ever_married', ylabel='count'>

People who are married have a higher stroke rate.

In [21]:
# Work Type & Stroke
sns.countplot(data=df,x='work_type',hue='stroke')
Out[21]:
<AxesSubplot:xlabel='work_type', ylabel='count'>

People working in the Private sector have a higher risk of getting a stroke. And people who have never worked have a very less stroke rate.

In [22]:
# Work Type & Stroke
sns.countplot(data=df,x='Residence_type',hue='stroke')
Out[22]:
<AxesSubplot:xlabel='Residence_type', ylabel='count'>

there not much difference in both attribute values. Maybe we have to discard it.

In [23]:
# Smoking & Stroke
sns.countplot(data=df,x='smoking_status',hue='stroke')
Out[23]:
<AxesSubplot:xlabel='smoking_status', ylabel='count'>
  • It is smal difference between who smokes and who does not smoke in regard to probability of getting stroke.
In [24]:
sns.boxplot(data=df,x='stroke',y='age')
Out[24]:
<AxesSubplot:xlabel='stroke', ylabel='age'>

Strokes are more likely to occur in people over 60.

Strokes are more likely to occur in people over 60. Some anomalies can be identified as strokes occurring in people under the age of 20. Given that our food and living habits have an impact on stroke, it's probable that the data is accurate. Another finding is that those over 60 years old make up the group of people who do not experience strokes.

In [25]:
sns.boxplot(data=df,x='stroke',y='avg_glucose_level')
Out[25]:
<AxesSubplot:xlabel='stroke', ylabel='avg_glucose_level'>

people having stroke have an average glucose level of more than 100.

In [26]:
sns.boxplot(data=df,x='stroke',y='bmi')
Out[26]:
<AxesSubplot:xlabel='stroke', ylabel='bmi'>

The relationship between BMI and the risk of stroke has not yet been clearly observed.

In [27]:
sns.countplot(data=df,x='smoking_status',hue='stroke')
Out[27]:
<AxesSubplot:xlabel='smoking_status', ylabel='count'>

there is not much difference in the chances of stroke irrespective of smoking status.

Since correlation check only accept numerical variables, preprocessing the categorical variables becomes a necessary step, we need to convert these categorical variables to numbers encoded to 0 or 1. We use labelEncoder from sklearn.preprocessing as it will be easy to decode a particular label back later after predicting if required.

In [28]:
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()

df['gender'] = le.fit_transform(df['gender'])
df['ever_married'] = le.fit_transform(df['ever_married'])
df['work_type'] = le.fit_transform(df['work_type'])
df['Residence_type'] = le.fit_transform(df['Residence_type'])
df['smoking_status'] = le.fit_transform(df['smoking_status'])


df.head()
Out[28]:
gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status stroke
0 1 67.0 0 1 1 2 1 228.69 36.6 1 1
1 0 61.0 0 0 1 3 0 202.21 28.9 2 1
2 1 80.0 0 1 1 2 0 105.92 32.5 2 1
3 0 49.0 0 0 1 2 1 171.23 34.4 3 1
4 0 79.0 1 0 1 3 0 174.12 24.0 2 1
In [29]:
corr = df.corr().round(2)
plt.figure(figsize=(10,7))
sns.heatmap(corr, annot = True, cmap = 'RdYlGn');
  • We can confirm that some of the variables are multicollinear based on the correlation matrix mentioned above. An example of this is the 0.68 correlation between the ever married and age columns.
In [ ]:

Data Processing & Modeling¶

All the predictor variables will be mapped to an array x and the target variable to an array y. The target variable is ‘stroke’ column.

In [30]:
# Models
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import SVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.cluster import KMeans
from sklearn import metrics
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
import plotly.figure_factory as ff
In [31]:
x=df.drop(['stroke'], axis=1).values  #Features
y=df['stroke'].values #target variable


# Evaluation
from sklearn.metrics import confusion_matrix, accuracy_score, classification_report

x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.3)
In [32]:
# show the value counts of the calsses in the target
# we can find data impalance
print(df['stroke'].value_counts())
df['stroke'].value_counts().sort_index().plot.bar()

0    4860
1     249
Name: stroke, dtype: int64
Out[32]:
<AxesSubplot:>

HANDLING IMBALANCED DATA¶

  • USING SMOTE TECHNIQUE TO HANDLE THE IMBALANCED DATA.
In [33]:
from imblearn.over_sampling import SMOTE

print("Before OverSampling, counts of label '1': {}".format(sum(y_train==1)))
print("Before OverSampling, counts of label '0': {} \n".format(sum(y_train==0)))

sm = SMOTE(random_state=2)
x_train_res, y_train_res = sm.fit_resample(x_train, y_train.ravel())

print('After OverSampling, the shape of train_X: {}'.format(x_train_res.shape))
print('After OverSampling, the shape of train_y: {} \n'.format(y_train_res.shape))

print("After OverSampling, counts of label '1': {}".format(sum(y_train_res==1)))
print("After OverSampling, counts of label '0': {}".format(sum(y_train_res==0)))

Before OverSampling, counts of label '1': 184
Before OverSampling, counts of label '0': 3392 

After OverSampling, the shape of train_X: (6784, 10)
After OverSampling, the shape of train_y: (6784,) 

After OverSampling, counts of label '1': 3392
After OverSampling, counts of label '0': 3392
In [ ]:

Model¶

In [34]:
def Model(model):
    model.fit(x_train,y_train)
    #Score 
    score = model.score(x_test, y_test)
    model_train_score = model.score(x_train, y_train)
    model_test_score = model.score(x_test, y_test)
    
    # Model Testing
    y_pred = model.predict(x_test)
    
    # Evaluation
    cm = confusion_matrix(y_test,y_pred)
    print(f'Accuracy Score: {accuracy_score(y_test,y_pred).round(3)}')
    
    conf_matrix = pd.DataFrame(cm,columns=['Predicted:0','Predicted:1'],index=['Actual:0','Actual:1'])
    
    z=cm
    fig = ff.create_annotated_heatmap(z, x=['Not Stroke','Stroke'], y=['Predicted Not Stroke','Predicted Stroke'],
                                      colorscale='Fall',xgap=3,ygap=3)
    fig['data'][0]['showscale'] = True

    fig.update_layout(title_text='<b>Confusion Matrix<b>',
                      title_x=0.5,
                      titlefont={'size': 15, 'family': 'Courier New'},
                      width=600, height=300,
                      xaxis_showgrid=False,
                      yaxis_showgrid=False,
                      yaxis_autorange='reversed', 
                      paper_bgcolor=None,
                      )

    fig.show()
In [35]:
# Decision Tree Classification
d_classif= DecisionTreeClassifier()
Model(d_classif)

Accuracy Score: 0.92
In [36]:
# Support Vector Machine Classification
RF_classif= SVC()
Model(RF_classif)

Accuracy Score: 0.958
In [37]:
# Random Forest Classification
RF_classif= RandomForestClassifier()
Model(RF_classif)

Accuracy Score: 0.957
In [38]:
# K Nearest Neighbour
RF_classif= KNeighborsClassifier()
Model(RF_classif)

C:\Users\drhem\anaconda3\lib\site-packages\sklearn\neighbors\_classification.py:228: FutureWarning:

Unlike other reduction functions (e.g. `skew`, `kurtosis`), the default behavior of `mode` typically preserves the axis it acts along. In SciPy 1.11.0, this behavior will change: the default value of `keepdims` will become False, the `axis` over which the statistic is taken will be eliminated, and the value None will no longer be accepted. Set `keepdims` to True or False to avoid this warning.

C:\Users\drhem\anaconda3\lib\site-packages\sklearn\neighbors\_classification.py:228: FutureWarning:

Unlike other reduction functions (e.g. `skew`, `kurtosis`), the default behavior of `mode` typically preserves the axis it acts along. In SciPy 1.11.0, this behavior will change: the default value of `keepdims` will become False, the `axis` over which the statistic is taken will be eliminated, and the value None will no longer be accepted. Set `keepdims` to True or False to avoid this warning.

C:\Users\drhem\anaconda3\lib\site-packages\sklearn\neighbors\_classification.py:228: FutureWarning:

Unlike other reduction functions (e.g. `skew`, `kurtosis`), the default behavior of `mode` typically preserves the axis it acts along. In SciPy 1.11.0, this behavior will change: the default value of `keepdims` will become False, the `axis` over which the statistic is taken will be eliminated, and the value None will no longer be accepted. Set `keepdims` to True or False to avoid this warning.

C:\Users\drhem\anaconda3\lib\site-packages\sklearn\neighbors\_classification.py:228: FutureWarning:

Unlike other reduction functions (e.g. `skew`, `kurtosis`), the default behavior of `mode` typically preserves the axis it acts along. In SciPy 1.11.0, this behavior will change: the default value of `keepdims` will become False, the `axis` over which the statistic is taken will be eliminated, and the value None will no longer be accepted. Set `keepdims` to True or False to avoid this warning.


Accuracy Score: 0.95
In [39]:
# Naive Bayes
RF_classif= GaussianNB()
Model(RF_classif)

Accuracy Score: 0.867
In [40]:
# Logistic Regression
lg_reg=LogisticRegression(max_iter=5000)
Model(lg_reg)

Accuracy Score: 0.958
  • From the above accuracy scores, Logistic Regression, Random Forest Classification and Support Vector Machine models all gives high accuracy score of 0.948.

conclusions¶

  • According to this study's findings, it is possible to forecast stroke prediction using machine learning algorithms. Well-known classification algorithms, such as Decision Tree, Support Vector Machine, K Nearest Neighbour, Logistic Regression, Random Forest Classification and Support Vector Machine were tested for their accuracy rates.
  • Not all strokes can be prevented by changing one's lifestyle. Even while some of these adjustments can greatly reduce your risk of stroke, they can all be highly advantageous. If the person immediately stops smoking, the risk of stroke will be lowered. There should be restrictions on alcohol.
  • Alcohol can raise blood pressure, which ups the chance of suffering a stroke. Maintain a healthy weight. Being overweight or obese puts a higher risk of having a stroke. Eating a balanced diet and exercising frequently will help maintain a healthy weight. It needs to take all these precautions to avoid a stroke.